In recent years, neural network models have been increasingly utilized to analyze the production behavior of oil and gas reservoirs, significantly aiding in the optimal design of fracturing parameters and enhancing production efficiency. However, the data-driven models established in current research only use well productivity, geological and engineering parameters, resulting in poor generalization ability. Moreover, under the data-lacking conditions, the performance of traditional neural network models is limited. To address these issues, this study proposes novel neural network models that integrate physical constraints and data augmentation. We introduce two models: the PC-ANN (Physical Constraints-Artificial Neural Network) for static productivity prediction and the LSTM-PC-ANN for time-series productivity prediction. These models incorporate physical constraints in their architecture and use generative adversarial networks (GAN) for data augmentation, enhancing prediction accuracy. Our findings show that the proposed models outperform traditional ANN and LSTM models, demonstrating a significant improvement in prediction accuracy and stability. The PC-ANN model, with its physical intermediate layer, reduces the root mean square error (RMSE) by 21.6% compared to the conventional ANN model. The LSTM-PC-ANN model, which integrates heterogeneous data, achieves a 52.2% reduction in RMSE over the standard LSTM model. Compared to the Arps Decline Model and the Autoregressive Integrated Moving Average (ARIMA) model, the RMSE can be reduced by up to 89.77%. These advancements indicate that the dual-driven approach, combining physical constraints and data augmentation, can effectively improve the prediction accuracy of horizontal well productivity.

The pore size in shale has reached the nanoscale. The phase behavior of fluids in nanopores deviates from classical theory, making conventional phase theory inapplicable to shale reservoirs. It is thus of great significance to clarify the phase behavior at the nanometer pore scale in shale. Based on the shale reservoir in the Kaybob of the Duvernay shale, Canada, 6 nm and 10 nm cubic model systems were constructed, and a microscopic physical model was prepared included pathways at 10 nm, 40 nm, 500 nm, and micrometer scale. The fluid phase behavior in nanopores was studied using molecular simulations and nanofluidic chip experiments. The results show that during the retrograde condensation process, the condensate gas in small-scale pores condenses first and is less likely to evaporate as the pressure decreases. The conclusions drawn are: (1) In nanopores, the dew point pressure of condensate gas is higher than that of the bulk phase, and the smaller the scale of nanopores, the higher the dew point pressure. (2) The volume ratio of condensate in nanopores is higher, increasing as pore size decreases. (3) The condensation rate of condensate in nanoporous media varies slowly with pressure drop; During evaporation, the condensation rate of condensate in small pores varies more slowly with pressure drop compared to the bulk phase but is faster in large pores.

The Duvernay shale gas condensate reservoir in Canada is a typical example of the application of staged fracturing technology in horizontal shale gas Wells. The productivity of staged fracturing horizontal Wells is closely related to sweet spot distribution and drilling technology. The main controlling factors affecting the distribution of sweet spots in the study area were identified, and different shale condensate zones were divided according to the distribution of hydrogen index HI and condensate ratio CGR. By analyzing the production curve of horizontal Wells that have been put into production, the declining law is determined, and the time and length of horizontal Wells are normalized to form the average production curve of horizontal length per thousand meters of a single well under the current technical conditions. Combined with the well spacing, horizontal section length, and fracturing design parameters of future horizontal Wells, the typical decline curve of gas and condensate oil in each zone is formed to predict the production of gas and condensate oil per well. The research results provide favorable support for development planning and scheme deployment.

The pore size in shale has reached the nanoscale. The phase behavior of fluid in nanopores deviates from classical theory, resulting in conventional phase theory no longer being applicable to shale reservoirs. It is thus of great significance to clarify the phase behavior at the nanometer pore scale of shale. Based on the shale reservoir in the Kaybob of the Duvernay shale in Canada, 6 nm and 10 nm cubic model systems were constructed, and a microscopic physical model was prepared included channels at 10 nm, 40 nm, 500 nm, and micrometer scale. The fluid phase behaviors in nanopores were studied through molecular simulation and nanofluidic chip experiments. Main results show that: (1) In nanopores, the dew point pressure of condensate gas is higher than that of the bulk phase, and the smaller the scale of nanopores, the higher the dew point pressure. (2) The volume ratio of condensate oil in nanopores is higher, and the volume ratio of condensate oil increases with the decrease of pore size. (3) The condensation rate of condensate oil in nano porous media varies relatively slowly with pressure drop; During the evaporation process, the evaporation rate of condensate oil in small pores varies more slowly with pressure drop compared to the bulk phase, while it is faster in large pores.

With the continuous improvement of shale oil and gas recovery technologies and achievements, a large amount of geological information and data have been accumulated for the description of shale reservoirs, and it has become possible to use machine learning methods for “sweet spots” prediction in shale oil and gas areas. Taking the Duvernay shale oil and gas field in Canada as an example, this paper attempts to build recoverable shale oil and gas reserve prediction models using machine learning methods and geological and development big data, to predict the distribution of recoverable shale oil and gas reserves and provide a basis for well location deployment and engineering modifications. The research results of the machine learning model in this study are as follows: ① Three machine learning methods were applied to build a prediction model and random forest showed the best performance. The R2 values of the built recoverable shale oil and gas reserves prediction models are 0.7894 and 0.8210, respectively, with an accuracy that meets the requirements of production applications; ② The geological main controlling factors for recoverable shale oil and gas reserves in this area are organic matter maturity and total organic carbon (TOC), followed by porosity and effective thickness; the main controlling factor for engineering modifications is the total proppant volume, followed by total stages and horizontal lateral length; ③ The abundance of recoverable shale oil and gas reserves in the central part of the study area is predicted to be relatively high, which makes it a favorable area for future well location deployment.

In order to overcome the defects that the analysis of multi-well typical curves of shale gas reservoirs is rarely applied to engineering,this study proposes a robust production data analysis method based on deconvolution,which is used for multi-well inter-well interference research.In this study,a multi-well conceptual trilinear seepage model for multi-stage fractured horizontal wells was established,and its Laplace solutions under two different outer boundary conditions were obtained.Then,an improved pressure deconvolution algorithm was used to normalize the scattered production data.Furthermore,the typical curve fitting was carried out using the production data and the seepage model solution.Finally,some reservoir parameters and fracturing parameters were interpreted,and the intensity of inter-well interference was compared.The effectiveness of the method was verified by analyzing the production dynamic data of six shale gas wells in Duvernay area.The results showed that the fitting effect of typical curves was greatly improved due to the mutual restriction between deconvolution calculation parameter debugging and seepage model parameter debugging.Besides,by using the morphological characteristics of the log-log typical curves and the time corresponding to the intersection point of the log-log typical curves of two models under different outer boundary conditions,the strength of the interference be-tween wells on the same well platform was well judged.This work can provide a reference for the optimization of well spacing and hydraulic fracturing measures for shale gas wells.

The proficient application of multistage fracturing methods enhances the status of the Duvernay shale formation as a highly esteemed shale reservoir on a global scale. Nevertheless, the challenge is in accurately characterizing unconventional fracture behavior and predicting shale productivity due to the complex distributions of natural fractures, pre-existing faults, and reservoir heterogeneity. The present study puts forth a Geo-Engineering approach to comprehensively investigate the Duvernay shale reservoir in the vicinity of Crooked Lake. To begin with, on the basis of the experimental results and well-logging interpretations, a high-quality petrophysical and geomechanical model is constructed. Subsequently, the establishment of an unconventional fracture model (UFM) takes into account the heterogeneity of the reservoir and the interactions between hydraulic fractures and pre-existing natural fractures/faults and is further validated by 18,040 microseismic events. Finally, the analysis of well productivity is conducted by numerical simulations, revealing that the agreement between the simulated and observed production magnitudes exceeds 89%. This paper will guide the efficient development of increasingly important unconventional shale resources.

The Duvernay Formation of Simonette Block is rich in shale oil and gas resources. The present-day in-situ stress field is a crucial parameter during the exploration and development in this block, a better understanding of which can assist with drilling and completion engineering, fracturing stimulation, and well deployment. In this study, the present-day in-situ stress in the Duvernay shale is predicted and analyzed based on well calculations and geomechanical modeling. The findings indicate that, I, NE-SW-trending is the dominant SHmax orientation. The SHmax and Shmin are the maximum and minimum principal stress, indicating that the Duvernay shale of Simonette Block is under a strike-slip faulting stress regime. II, The present-day in-situ stress in the Duvernay shale is heterogeneously distributed, which is controlled by lithology difference, fault development, and distribution. High stress values in the upper shale member D layer are mainly in the northwestern and central parts of Simonette Block. III, The present-day differential stress in the Duvernay shale mainly ranges from 15 MPa to 30 MPa. Bedding-parallel fractures are relatively developed, aiding in the creation of complex fracture networks. The results are expected to offer geological references for further development of shale oil and gas in the Simonette Block.

Plunger lift is one of the commonly used processes for liquid extraction and gas production in gas wells. Owing to the gap between the plunger and the tubing wall, liquid leakage and gas channeling during the lifting process are inevitable. Liquid leakage reduces the amount of lifted liquid, and gas channeling is not conducive to the complete utilization of formation energy. Furthermore, the gas that breaks through the gap into the top of the plunger also hinders plunger movement. To address this problem, we used CFD to simulate plunger movement in the tubing and analyzed the influence of various factors, such as plunger velocity, pressure difference between the upper and lower ends, and plunger diameter, on liquid leakage and air channeling. The results indicated that when the plunger rises and drains the liquid, the plunger velocity increases, the leakage between the plunger and the tubing increases, and the gas channeling volume decreases. When the pressure difference between the upper and lower ends of the plunger increases, leakage loss is reduced. The larger the plunger diameter, smaller the interval between the plunger and the tubing, and smaller the less leakage loss. The results of the study validate the influencing factors of plunger lift leakage and gas channeling, optimizing the plunger structure and improving the efficiency of plunger lifting.

Well structures with ultra-long sections have become one of the most applied technologies in the field of shale gas development. While there have been many technical challenges, enhancing the breaking efficiency and stability of polycrystalline diamond compact (PDC) bits has become an essential issue of focus. Since 2013, the well structure in the Duvernay area has been optimized multiple times, and the rate of penetration (ROP) of the entire wellbore has nearly doubled. However, there are significant differences in terms of the performances of different PDC bits, and there is still room for improvement to optimize these drill bits. For this reason, a confined compressive strength test was conducted to obtain the rock mechanical parameters from shale cores extracted from the long horizontal section. Using these data, a finite element model (FEM) was developed with a corresponding scale. A calibration of the elastic-plastic damage constitutive models was then performed using the FEM. The breaking mechanism of three different PDC bits was examined using a “PDC bit-bottom hole” interaction FEM model, facilitating guidance for bit selection and design optimization: (1) The type B PDC bit, which has four blades and 20 cutters, exhibited the highest mechanical specific energy (MSE) and the lowest vibration across three directional mechanical characteristics. This design is recommended for engineering applications. (2) Lower axial vibrations were produced when the CDE was used as the rear element when compared to those when using the BHE. However, an increase within an acceptable range was observed in the TOB and circumferential vibrations. Thus, for redesigning work on the type B bit, the assembly of the CDE is suggested. (3) A decrease in the MSE and vibration in three directional mechanical characteristics was observed when the depth of cut (DOC) was varied between 1.5 and 2.0 mm. A broadening in the range of lateral forces was noted when a DOC of 2.0 mm was used. Therefore, for the redesign of the type B bit, the assembly of CDEs as rear elements at a DOC of 1.5 mm is recommended. In conclusion, a new practical method for the selection and optimization of PDC bit design, based on rock mechanics and the FEM theory, is proposed.

In this paper, taking Block G in Canada as an example, combined with the data of the working area, the Pearson–MIC comprehensive evaluation method was adopted to optimize the key parameters of productivity. Based on the analytic hierarchy process, the weight of each parameter was calculated, the grade of evaluation index of the “sweet spot” was divided, the standard of the sweet spot was established, and the distribution of the superimposed sweet spot was finally depicted. The results show that lateral length, number of stages, volume of fluid, and amount of proppant are the key engineering parameters of horizontal well, and lateral length is an independent key engineering parameter. The cumulative gas production in the first two years was normalized on the lateral length to eliminate the engineering influence, and the total organic carbon (TOC) was finally determined as the key geological parameter, whereas porosity and water saturation were the secondary key parameters. The area of Type I sweet spots accounts for 24.2% in the Series Upper and 23.1% in the Series Lower. This study proposed a new sweet spot prediction idea based on the influence of geological factors on productivity, and its results also laid a foundation for the subsequent placement of horizontal wells in Block G.

The shallow Montney reservoir in S area of Western Canadian Basin has been proved to be proliferous by the production performances of new horizontal wells, so it is urgent to characterize the reservoir distribution. But, both the limitation of well data and poor seismic vertical resolution bring challenges to the prediction of this new potential layer. A novel workflow, based on high resolution processed 3D seismic, logging and core data, integrating multi-seismic attribute inversion with minimum support algorithm, logging interpretation and 3D geological modeling, was introduced to fix the challenge. First, choose an objective log which reflects lithology and physical properties variations. Through multi-mineral logging interpretation and cross-plotting, Gamma ray was optimized; second, improve the vertical resolution of original seismic data by sparse layer reflectivity inversion method, before inversion. This process is very critical and guarantees the whole seismic attributes inversion is conducted based on high resolution data; third, extract seismic attributes from processed seismic volume and analyze their sensitivity to lithology variations, providing basis for multi-seismic attributes combinations. Fourth, link the Gamma ray of key wells to high resolution seismic data, and invert the log (Seismic GR) by multi-seismic attributes combinations based on minimum support algorithm (MSA). Compared to traditional PPN (probabilistic neural network), this method is more stable in areas with less wells. Finally, resample the Seismic GR volume into static model, and calibrate it with GR of all wells and P-impedance volume, this step improve the prediction accuracy of inter-wells and resolution in vertical, greatly reducing the uncertainty of lithology prediction between wells. This workflow enables the well resolution to match with the processed seismic resolution, reduces the instability of multi attributes inversion resulting from limited well training data. Besides that, after the validation of well logs and P-impedance, the multi-seismic attributes inversion results (Seismic GR) are more accurate and with high vertical and lateral recognition. The Seismic Gamma ray shows consistency of Montney reservoir delineation, the more developed the reservoir, the smaller the Seismic GR values, with threshold value of favorable lithology from wells, it is easy to delineate reservoir distribution of Montney formation in S area with constraint of lithology interpretation results from logging.

The dynamic productivity prediction of shale condensate gas reservoirs is of great significance to the optimization of stimulation measures. Therefore, in this study, a dynamic productivity prediction method for shale condensate gas reservoirs based on a convolution equation is proposed. The method has been used to predict the dynamic production of 10 multi-stage fractured horizontal wells in the Duvernay shale condensate gas reservoir. The results show that flow-rate deconvolution algorithms can greatly improve the fitting effect of the Blasingame production decline curve when applied to the analysis of unstable production of shale gas condensate reservoirs. Compared with the production decline analysis method in commercial software HIS Harmony RTA, the productivity prediction method based on a convolution equation of shale condensate gas reservoirs has better fitting affect and higher accuracy of recoverable reserves prediction. Compared with the actual production, the error of production predicted by the convolution equation is generally within 10%. This means it is a fast and accurate method. This study enriches the productivity prediction methods of shale condensate gas reservoirs and has important practical significance for the productivity prediction and stimulation optimization of shale condensate gas reservoirs.

Tight gas reservoirs are mainly developed by multistage hydraulic fracturing horizontal wells (MSHFHWs). A type well provides average production profiles based on real well data and can be constructed from multiple wells to investigate the behavior of the reservoir. For unconventional reservoirs, type wells are the key to reserve calculations and medium- and long-term field development planning. Both geological and completion parameters are key factors affecting single well performance of MSHFHWs. Based on the drilling, hydraulic fracturing, and production data for over 1,800 MSHFHWs in the Montney tight gas reservoir in the Groundbirch region of the Western Canada Sedimentary Basin (WCSB), the main hydraulic fracturing factors affecting the production performance of MSHFHWs were investigated. A rate transient analysis- (RTA-) assisted workflow for type well construction is proposed based on existing production data and considering the geological and engineering factors. Based on the field data, the main hydraulic fracturing factors that affect the production performance of the MSHFHWs in Montney are lateral length, proppant tonnage, and the number of stages. Base type wells are predicted from the P50 wells, which are selected from the wells with normalized lateral length and the same fracturing technique and proppant tonnage. The base type well represents the well performance for a specific drilling and completion background. RTA was introduced to scale up the base type well to predict the type well of new completion design. The new workflow predicts both the base type well with a specific drilling and completion background and the upgraded type well, which uses new completion design. It is highly meaningful and provides a valuable reference to practical studies involving type well prediction in unconventional gas reservoirs.

The recent remarkable increase in induced seismicity in Western Canada has been largely attributed to hydraulic fracturing in unconventional reservoirs. The nucleation of large magnitude events has been demonstrated to be closely linked to site-specific geological and operational factors. A mitigation strategy of fracturing-induced seismicity concerning both factors has not been well investigated. In this paper, a comprehensive investigation of risk mitigations from induced seismicity is conducted based on the formation overpressure, distance to Precambrian basement, proximity to faults, fracturing job size and safe hydraulic fracture-fault distance. It is found that the middle-south region near Crooked Lake is an optimal region for fracturing operations with low formation pressure, a great distance to the basement and relatively fewer pre-existing faults. A field case study suggests that fracturing operations of three new horizontal wells are successful with low magnitude induced events and with high production performance, demonstrating the applicability of a comprehensive approach of seismicity risk mitigations. Such an approach can be applied to other field cases to mitigate the potential fracturing-induced seismicity in unconventional reservoirs.

In consideration of vertical formation heterogeneity, a basic nonlinear model of 1D commingled preferential Darcian flow and non-Darcian flow with the threshold pressure gradient (TPG) in a dual-layered formation is presented. Non-Darcian flow in consideration of the TPG happens in the low-permeability tight layer, and the Darcian kinematic equation holds in the other high-permeability layer. The similarity transformation method is applied to analytically solve the model. Moreover, the existence and uniqueness of the analytical solution are proved strictly. Through analytical solution results, some significant conclusions are obtained. The existence of the TPG in the low-permeability tight layer can intensify the preferential Darcian flow in the high-permeability layer, and the intensity of the preferential Darcian flow is very sensitive to the dimensionless layer thickness ratio. The effect of the layer permeability ratio and layer elastic storage ratio on the production sub-rate is more sensitive than that of the layer thickness ratio. In addition, it is strictly demonstrated that moving boundary conditions caused by the TPG should be incorporated into the model. When the moving boundary is neglected, the preferential Darcian flow in the high-permeability layer will be exaggerated. Eventually, solid theoretical foundations are provided here, which are very significant for solving non-Darcian seepage flow problems in engineering by numerical simulation validation and physical experiment design. Furthermore, they are very helpful for better understanding the preferential flow behavior through the high-permeability paths (such as fractures) in the water flooding development of heterogeneous low-permeability reservoirs; then, the efficient profile control technology can be further developed to improve oil recovery.

Well spacing is one of the key factors affecting the economic performance of shale gas reservoirs with horizontal wells. In this paper, based on the analysis of dynamic data of 180 wells in D shale gas field, we have picked the characteristic curve to identify the boundary feedback signal, which can be used to identify the inter-well interference (IWI) of different well spacing. In addition, we have innovatively formed a set of methods and processes to utilize multi-information such as micro-seismic, tracer, and adjacent well dynamic interference to analyze quantitatively the degree of IWI, which can be used to quantify the degree of IWI of different well spacing. Finally, based on the principle of keeping proper IWI and combining with the numerical simulation, the optimal well spacing is determined. Pick the characteristic curve to identify IWI. By analyzing the producing characteristics of 180 individual horizontal wells of D shale gas reservoir, we pick the double logarithmic curve of production and time as the characteristic curve that can clearly identify the boundary feedback signal. Based on this, we contrastively analyze the characteristic curve of different well spacing and find that the well spacing below 200 m reaches the boundary much earlier than the 400 m well spacing, which shows that there is serious IWI in the well spacing below 200 m. Utilize multi-information such as micro-seismic, tracer, and adjacent well dynamic interference to quantify the degree of IWI of different well spacing and determine optimal well spacing. For the micro-seismic information, we found that the micro-seismic events of each fracturing stage distributed normally along the vertical direction of horizontal well. Base on it, we standard the cumulative probability distribution curve of the micro-seismic events of each stage, and obtain the typical cumulative probability distribution curve. Then, we use the typical cumulative probability distribution curve to simulate IWI by arranging the curve according to a certain well spacing. Furthermore, we calculate the proportion of the acreage of the overlapping area of two or more curves in the middle well to the acreage of the whole individual curve to quantify the degree of IWI. For tracer information, we found that the farther away from the injection well, the less the tracer flow-back. We normalize the tracer flow-back concentration ratio (TFBCR) of injection well, and then use the TFBCR of the wells at different well spacing from the injection well to quantify the degree of the IWI. For the adjacent well dynamic interference information, we found that the smaller the well spacing, the greater the change in production caused by the shut-in of adjacent wells, the greater the difference in production between side wells and non-side wells in the same well pad. Base on it, we calculate the average production change ratio caused by shut-in of adjacent wells and the production difference ratio between side wells and non-side wells to quantify the degree of IWI. Combining the above information, the range of the degree of IWI for different well spacing can be quantitatively determined. Then, based on the principle of keeping proper IWI and combining with the numerical simulation, the optimal well spacing is determined. We applied the research achievements of this paper in D shale gas field, and found that the well spacing below 200 m had occurred serious IWI, the degree of IWI intensity of the 200 m well spacing was 11%–30%, that of the 300 m well spacing was 5%–14%. Then, based on the principle of keeping proper IWI and combining with the numerical simulation, we determined that the optimal well spacing of D shale gas field was 300 m. Consequently, we suggested expanding the well spacing of D shale gas field from 200 m to 300 m. After implementation, the individual well performance of 300 m well spacing had increased by 40% than that of 200 m well spacing and about 200 wells would be saving in the FDP.

AbstractDuvernay shale spans over 6 million acres with a total resource of 440 billion barrels’ oil equivalent in the Western Canada Sedimentary Basin (WCSB). The oil recovery factors typically decrease with the decreasing of gas-oil ratio (GOR) in oil window of Duvernay shale. The volatile oil recovery factors are typically 5–10%. Enhanced oil recovery technologies should be applied to improve the economics of the reservoirs. In this paper, the volatile oil from the Duvernay shale was taken as an example for phase behavior study. We analyzed the nanopore confinement on phase behavior and physical properties of Duvernay shale oil. The shift of critical properties was quantified within nanopores. With the confinement of nanopores, the viscosity, density, and bubble point pressure of the oil decrease with the shrinking of the pore size. Minimum miscibility pressure (MMP) was calculated for different injected gases. The MMP from high to low is N2>CH4>lean gas>rich gas>CO2. In the case of injecting the same gas component, the MMP decreases as the pore size decreases. The wellhead rich gas is suggested to be the main gas source for gas injection in Duvernay shale. The formation pressure should be rapidly increased to the MMP and maintained close to it, which would help to improve the effect of gas injection and enhance shale oil recovery. This paper can provide critical insights for the research of shale oil gas injection for enhanced oil recovery.

Duvernay shale is a world class shale deposit with a total resource of 440 billion barrels oil equivalent in the Western Canada Sedimentary Basin (WCSB). The volatile oil recovery factors achieved from primary production are much lower than those from the gas-condensate window, typically 5–10% of original oil in place (OOIP). The previous study has indicated that huff-n-puff gas injection is one of the most promising enhanced oil recovery (EOR) methods in shale oil reservoirs. In this paper, we built a comprehensive numerical compositional model in combination with the embedded discrete fracture model (EDFM) method to evaluate geological and engineering controls on gas huff-n-puff in Duvernay shale volatile oil reservoirs. Multiple scenarios of compositional simulations of huff-n-puff gas injection for the proposed twelve parameters have been conducted and effects of reservoir, completion and depletion development parameters on huff-n-puff are evaluated. We concluded that fracture conductivity, natural fracture density, period of primary depletion, and natural fracture permeability are the most sensitive parameters for incremental oil recovery from gas huff-n-puff. Low fracture conductivity and a short period of primary depletion could significantly increase the gas usage ratio and result in poor economical efficiency of the gas huff-n-puff process. Sensitivity analysis indicates that due to the increase of the matrix-surface area during gas huff-n-puff process, natural fractures associated with hydraulic fractures are the key controlling factors for gas huff-n-puff in Duvernay shale oil reservoirs. The range for the oil recovery increase over the primary recovery for one gas huff-n-puff cycle (nearly 2300 days of production) in Duvernay shale volatile oil reservoir is between 0.23 and 0.87%. Finally, we proposed screening criteria for gas huff-n-puff potential areas in volatile oil reservoirs from Duvernay shale. This study is highly meaningful and can give valuable reference to practical works conducting the huff-n-puff gas injection in both Duvernay and other shale oil reservoirs.

In recent years, Unconventional gas reservoirs development in North America has been in full swing, which have become one of the main sources of natural gas in North America. It is well known that unconventional gas reservoirs with very low permeability must be artificially fractured to get commercial/industrial flow. Multi-stage fractured horizontal Wells based on artificial fractured are the main well type of unconventional gas reservoir. The artificial fracturing technology seriously affects the well performance of the Multi-stage fractured horizontal well. However, even if the same fracturing technology is used, the performance of wells in the same reservoir is still very different. It is shown that the factors influence the well performance of unconventional gas reservoirs are not only the external factors, such as lateral length, stages, stage spacing, and so on, but also the internal factors, such as burial depth, net pay, porosity, and so on. When all kinds of factors are mixed together, which factors are the key parameters that influence the performance of the Multi-stage fractured horizontal Well of Unconventional gas reservoir? Due to the poor correlation of the traditional single factor analysis method, this paper uses the combined traditional single factor analysis with outlier analysis method to analyze the influence of various parameters on well performance. Firstly, based on the traditional cross-plotting between parameters and production index, combined with practical knowledge, get the hypothetical correlation between each parameters and production index. Meanwhile, according to the distribution of each parameter in all statistical wells, the P50 value of each parameter is determined. Then, pick several wells with the very best and very worst production indexes, to prove, or disprove, hypothetical correlation between each parameter and production indexes by looking for opposite signatures in the very best and very worst wells. The more the matched signatures, the stronger the parameter influence on well performance. Finally, the key parameters can be find out by analyzed the matched signatures. Taking the Duvernay shale reservoir which has 150 producing wells as an example and using the first year cumulative production as the production index, this paper selects the very best production index 5 wells and the very worst 5 wells to analyze the influence of each parameter on the production index based on the above method. Finally, the key parameters of Duvernay shale reservoir are find out. The key reservoir parameters are CGR, gas saturation, TOC content and pressure coefficient. The key fracturing parameters are lateral length, proppant per stage, fluid per stage, proppant concentration and flow-back rate. The determination of the key reservoir parameters provides the basis for the prediction of the sweet point area, and the determination of the key fracturing engineering parameters can guide the optimization of the fracturing process.